Group delay perception thresholds depend strongly on signal content and context

Evidence

At best, our hearing can detect group delay changes, which are smooth in time, of 1 ms within a critical band. For example, sound signals from the elements of a multi-way loudspeaker at different distances from the listener arrive at different times. In laboratory conditions, delay differences of a little below 1 ms can be detected when directly compared to the nondelayed sound. This means that the distances from the elements of a loudspeaker to the listener should be similar, not deviating by more than 20–30 cm, which is normally not a problem in loudspeaker design. … Group delay changes have to be about 3–5 ms with speech and music before they can be perceived. If the changes are of the order of 5–10 ms, they begin to be detectable if the effect of the listening room is minimal. With some phase-insensitive signals, such as noise or diffuse sound, changes in group delay may not be perceived even with longer delays. … Although the changes in group delay may be perceived as changes in the timbre of sound, the effect on communication, or on speech intelligibility, can be small. For example, even 100- ms group delay changes can be introduced into speech signals before intelligibility starts to vanish. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1280

Categorization

• XREF: Connects to everyday audio engineering constraints on loudspeaker element placement; a clean design rule (elements within 20-30 cm path difference) an engineer could apply directly.